What’s the difference between you and a very powerful computer? This is the question posed by artificial intelligence research, by Tom Stoppard’s new play The Hard Problem, and by Friedrich Hayek in a classic 1945 paper in The American Economic Review, The Use of Knowledge in Society. To many people, including Stoppard’s heroine, the differences are obvious: emotion, the capacity for moral reasoning, free will perhaps. These features make us human, and it is our human needs that markets and market information in principle serve.
In his paper, Hayek asks whether a central planner could organise the use of resources in society better than the disorganised, decentralised market system, which can be overwhelmed by emotion and irrationality. He argued that no planner could ever be intelligent and fast enough to use the vast array of changing and often contradictory pieces of information about society. “We must look at the price system as such a mechanism for communicating information,” he wrote. The economic (never mind political and moral) disaster of centrally planned economies vindicated his argument. As many people have since pointed out (like Paul Seabright in his wonderful book The Company of Strangers) there is something rather magical about how well markets co-ordinate the demand and supply of so many goods between so many millions of people, using the signals sent by prices.
Today’s information and communication technologies offer huge potential for making markets more efficient at this matchmaking process by quickly disseminating price signals. One now-classic study was Robert Jensen’s investigation of fish prices in Kerala, India before and after the introduction of mobile phones along the coast. The ability to access and convey information led to a dramatic convergence of fish prices between different markets. Fishermen’s incomes rose. Consumers gained (a little) too: not only did the fish go to the markets where it was most highly valued, but there was also less wastage. Other studies of mobiles and agricultural prices confirm that, where producers can act on the improved price information, the new technology improves economic efficiency.
So is it possible to imagine that a modern central planner, a sufficiently powerful computer with vast access to information at its disposal and rapid processing capacity, could achieve the same efficiency as the decentralised markets? High-frequency trading (HFT) in modern financial markets might be close to this information-rich, super-efficient state, trading on tiny titbits of information at nearly the speed of light. But it is hard to feel confident that this is improving efficiency in the same way as getting better crop price information to low-income farmers in India or Niger.
Roughly half of equity trades in the US and UK financial markets are now carried out by these ultra-fast thinking machines. HFT involves computers trading securities according to algorithms, with no additional human input, raising the tantalising thought that a network of computers transacting so fast could act as a virtual central planner. There is evidence though that the trading is characterised by many ‘flash crashes’, like that of May 2010, although most of them are over so quickly that nobody notices. A commonly shared suspicion of HFT was examined in Michael Lewis’s book Flash Boys: if it’s worth many hundreds of millions of dollars to invest in ever-faster communications networks for a few milliseconds’ worth of advantage, is that a measure of how well they are fleecing the everyday investor?
It is hard to resist the thought that high-frequency trading is the culmination of the ‘performativity’ some social scientists believe characterises finance. The linguistic philosopher J.L. Austin coined this word to refer to phrases such as “I name this ship the QEII”: to say the words is to perform the act. Donald Mackenzie has argued that the options pricing model of financial economics is performative because it brought into existence modern options markets. Before the model, nobody really knew how to price options, and the market was minuscule. With HFT, has economic theory – specifically the Efficient Markets Hypothesis – brought into existence the rational, calculating, well-informed agents it assumes? If so, why is it not obviously more efficient to have markets composed of algorithms rather than people?
Hayek would not have been surprised by the general suspicion. He wrote, ”To gain an advantage from better knowledge of facilities of communication or transport is sometimes regarded as almost dishonest, although it is quite as important that society make use of the best opportunities in this respect as in using the latest scientific discoveries.” In other words, arbitrage is a socially useful function even though people tend to believe it is a bit dodgy. So perhaps the algorithms are worhwhile despite our suspicion of HFT?
I’m not so sure. The greater and faster information accessible via mobile phones in agricultural production differs qualitatively as well as quantitatively from the greater and faster information the HFT algorithms are providing via computerised prices. Information is about something, a signal; in the case of food markets, it is about what foodstuffs people want to buy and what is available for sale. If anything, better information in these markets reduces the need for arbitrage and liquidity; the matching of supplies and demands is improved without intermediation.
It isn’t clear that there are human wants behind the nearly-light-speed financial trading of HFT. The defence of HFT is that computerised trading increases liquidity, but liquidity is only necessary when it is hard to match demand and supply. If the matching process works well, more liquidity might not be useful. And in any case, many humans are troubled by HFT’s disconnect from the human desires that financial markets should ultimately serve.